A Beach-Read Beatdown
I thoroughly enjoyed a recent opinion essay in The New York Times by Curtis Sittenfield, in which she went head to head with ChatGPT in a writing contest to see which could write the better beach read. This type of contest is just the latest gimmicky piece of content created to feature the writing of a chatbot for the express purpose of letting readers assess it in context.
By
The elements for the prompts were voted on by NYT readers, and the results were pretty damning.
They don’t put a thumb on the scale: The writing samples are anonymized and readers are allowed to judge for themselves.
I don’t think it takes a deeply discerning person to be able to pick out the one written by a human. Where it sings, the other is just serviceable.
It’s not a technical deficiency that gives away which is which. It’s soul. One has it, and the other, at best, is attempting it (though we know it’s not “attempting” anything at all).
What made me happy was how sure I was reading Sittenfield’s that I’d identified the human example. It didn’t have to wonder. It was alive (and a damn good story). The other just wasn’t.
Is that enough to make me feel better about artificial intelligence coming for my job as a writer? No, not really. Because too often readers are perfectly content with “serviceable” and “good enough.” But does it give me a smug bit of satisfaction that humans still solely posses the inputs needed to make a story feel alive such that a reader can confidently know, “This is the one”? Yeah, it does.